MRSA spread in the ICU is related to nursing workload

Excerpt

This is an interesting idea for a study but I do not believe that the conclusions in the discussion are supported by the results as there are major problems with the methodology. Firstly, when mean, peak and trough levels of the indices are considered there are 26 possible correlations, of which 10 are reported as statistically significant. What conclusion is it safe to draw when, for example, the peak and trough staffing levels are weakly correlated with the number of MRSA cases but the mean staffing level apparently isn't? Secondly, there are fewer significant correlations of potential transmissions (three) than of actual cases (seven) with nurse staffing indices. Given the likely delay between the time of cross-infection and subsequent colonisation (sufficient to be detectable by surveillance swabs), this seems difficult to explain. Thirdly, and most importantly, the authors seem to have made the common error of confusing correlation with causation. It is interesting that MRSA cases and potential transmissions are raw figures — i.e. not related to the number of patients present. It is likely that nurse staffing indices will be most extreme when the ICU is full. MRSA cases and cases of potential transmission will clearly also be higher when there are more patients. (Similarly a 300-bedded hospital would expect fewer MRSA cases than a 1000-bedded hospital.) This is a plausible explanation for the weak correlations reported in the study. …